In this paper, we develop an optimization-based systematic approach for the
challenging, less studied, and important problem of optimal partitioning of
multi-thermal zone buildings for the decentralized control. The proposed method
consists of (i) construction of a graph-based network to quantitatively
characterize the thermal interaction level between neighbor zones, and (ii) the
application of two different approaches for optimal clustering of the resulting
network graph: stochastic optimization and robust optimization. The proposed
method was tested on two case studies: a 5-zone building (a small-scale
example) which allows one to consider all possible partitions to assess the
success rate of the developed method; and a 20-zone building (a large-scale
example) for which the developed method was used to predict the optimal
partitioning of the thermal zones. Compared to the existing literature, our
approach provides a systematic and potentially optimal solution for the
considered problem.